Skip to main content

Full text of "mit :: whirlwind :: M-series :: M-1347 Discussion of Magnetic Drum Systems at Engineering Research Associates 27-28 November 1951"

See other formats

Memorandum M -134-7 

Page 1 of 7 

Ligital Computer Laboratory 
Massachusetts Institute of Technology 
Cambridge, Massachusetts 


To: ' J. W, Forrester 

From: E. S. Rich 

Late: Lecember 7, 1951 

Abstract: Representatives of this laboratory visitec Engineering Research 
Associates in St. Paul for a two day discussion on magnetic 
drums. An outline of the status of the work done by E.R.A. 
indicates that satisfactory progress is being made. The buffer 
drum system was discussed in considerable detail and several 
major changes were requested in order to permit the use of 
this drum in our applications. Satisfactory agreements were 
worked out on these points and E.R.A. will revise their plans 
accordingly. Some details of circuitry were discussed. They 
have plans for using ferrite material in the cores of their 
magnetic heads. Final decision on whether this material can 
be uped will be made after 3 1. is. learned whether the desired 
core shape can be obtained by a molding process rather than 
by machining. The use of ferrite core materials promises an 
attractive simplification in their writing circuits. A schedule^ 
for breakdown of power vdring to provide marginal checking 
facilities was also worked out. 

The following men were present for a discussion on magnetic 
drums at the Engineering Research Associates, St. Paul, on 27-28 
Noyember, 1951:- from E.R.A., Jack Hill, Bill Butler ano Bob Eulbergj 
fripm M.I.T., Taylor, Vdeser, O'Brien, Walquist and Rich. The principal 
it^ms discussed were the present status of their work, the changes in 
the systems which we thought necessary, and various details of circuitry. 



,1 I 

1. Fabrication of the Magmetic Lrums . They ©intimate the magnetic 
dn^ms to be abciut B0% complete. The work which remaijns to be done includes 
machining the aifffSces of the dniras, boring^holes in ^the jacket for the 
ma^etic heads, ^ and reaming for the bearings. 



Memorandiim M-13A7 

2. Electronic Componentg . They have made a rough estimate of the 
electronic components -which they are sure vdll be needed and have placed 
orders for those , particularly for items vMch they believe -will be 
difficult to obtain. The parts ordered comprise approximately 8O5S of 
the total components needed. Remaining components vdll be ordered v;hen 
a more detailed analysis of circuit design has been made, Ydth present 
delivery schedules they do not expect to be able to obtain ail the needed 
materials before July, 1952, 

Crystal diodes were the one item which appears to present a 
bottleneck at, the present time. They had decided to use.Sylvania glass > 
crystals but learned that the type they want (IN56A) is virtually un- 
available. They have General Electric as an alternate supplier, although 
no order has been placed with them yet. Vie suggested that they contact 
Kemtron for availability of the crystals they need. It appears that they 
have little success in previous dealings with Kemtron, but we indicated 
that we could probably assist them in obtaining needed crystals from 

3* Power Supplies , 'Jhey plan to use motor generator sets for 
the d-c sujjplies and have placed orders for these. They had a question 
on what sort of supplies would be used for filament power, VJe informed 
them that within a week or two we would know whether we will have a 
filament supply of sufficient capacity to handle their drums. As a 
safety measure,, hov;ever, they are planning to order an Inductrol regulator 
system from General Electric Company which can be used to supply regulated 
filament power. This unit costs in the neighborhood of ^800 and therefore 
there would no hesitation to cancel the order Icter if necessary. They 
inquired about v;hat type of priraaiy power we can supply and were told that 
a 230 volt, 3 -phase, 3 vdre system would be the best choice/ \'5e agreed 
to make a week's recording of the voltage variations that occur on this 

4* Cabinets . They have ordered materials for the metal cabinets 
for the equipment. Those cabinets are to be 2iJ" wide and 81" high and 
have been oi'dered in three 10' sections. It is estimated that the two 
drum systems will occupy approximately 12* and 18' of cabinet space 
respectively, Tj[iis amount of cabinet space inciucec a safety factor so 
tha-b it would take care of expansion of the systems at a later date. 
The phanges in the buffer drum system which we requested will require 
some of this extra space. It is planned to mount the mdgne.tic drums 
within these cabinets along with the circuits so that the only imit which 
must be located separately is the motor generator set. It is their plan 
to have all wiring to these drums come through the top of the cabinets. 
They estimated that approximately 7 k.w, of heat would be displaced in 
the ^vm systems so that we should probably plan for at least 10 k.w. of 
air ponditioning capacity. 


Memorandum M -13-47 

5« LraiTJnga , They estimated tha\: their schematic draxTings Y^^ero 
ahcut SO^ complete. These drawings are suggested layout cravings of 
standard circuits on their standard chassis. A considerable amount of 
this v;ork v/ill have to be redone as a result of changes '\7laich we requested 
during our visit. 

6. Magnetic Heads . At the present time they are hopeful of being 
able to change their head design ii-om the use of metal core material to 
the use of ferrite cores. The use of the ferrite material, since it has 
a higher incremental permeability, permits one to ct^t dovm the I£!F needed 
to accomplish recording, and their experience indicates that the use of 
such heads in conjunction with a red oxice coating on the drum will permit " 
them to use a much simpler circuit. Specifically, they feel that this 
reduction will permit them to use a type 7AK7 tube for the recording circuit 
and thus combine the amjxLifier and gcte« The biggest question on whether 
ferrite material can be used is a question of ffbri eating the cores. The 
experimental units which they have used to date were obtained by machining. 
There is some question whether they can be fabricated simply by a molding 
process. They hope to obtain an answer to this question by January 15, 
1952 and therefore are not going to make a decision about the recording 
circuits until that time. Hill stated that ho ?;ould place an order for 
sufficient ferrite cores to be shipped as part of an experimental group 

in July of next year. If these cores did not prove to bo satisfactory 
it would be possible to obtain the necessary cores by machining, although- 
at a considerable expense. 

7. Packaging. They plan, to use a new type of plug-in chassis ' 
which has been developed by another group at E.R.A. Principal changes 
in this new chassis are provision for using regular size tubes as well 
as miniature tubes, iilso they are planning to use a screw type lug for 
mounting crystals to avoid soldering. 

They have decided to discard the use of the type 5687 tubes 
and have chosen types 6AV5 (G.E.) and 6AU5 (Sylvania) as substitutes for 
hard tube writing circuits. These tubes would not be necessary for the 
single heads if they can use ferrite core materials, but they will be 
necessary for use v;ith the dual heads since these can not be readily 
adapted for the ferrite cores. 


A considerable portion of the time during our visit ^vas spent 
in discussing the block diagram layout of the buffer drum. The changes 
which we requested were the result of our stuty of the intended appli- 
cations of this drum. The following changes in the system were agreed 
upon: [ 



Memorandum M-13^7 

1. Angular Position Counter . They \jill design the flip flops 
of the counter so that output signals can he obtained to drive a matrisc 
smtch. Furthermore they will provide a matrix on three of the flip 
flops so that an 8 position output svdtch will he included. The outputs 
of this svdtch will be capable of driving two gate tubes each. 

2» Coinci pence Dete cto r. They have decided to use 7AK7 gate 
tubes for the coincidence detector and heo been planning to mix the 
outputs of all 11 digits into a single inverter-gate tube combination. 
\ie requested that they provide for three different coincidence pulses 
so they will separate the coincidence gates in such a manner that mixing 
will be done to three inverter-gate tube combinations to give the three 
coincidence pulses which we want« 

3. Storage Address Register . The storage address register will 
be wired so that the left 11 digits can be cleared independently of the 
right /^ digits. We found out that they plan to construct one digit each 
of the angular position counter, the coincidence detector, and the storage 
address register on a single chassis. Actually, tv/o such digit columns 
will be built on a chassis but will be independent of each othero The 
interlace between the storage address register and the angular position 
counter can be obtained by appropriate interconnection of the digits of 
the address register and of the counter. 

A» Heading Amplifiers . They considered that it would be impossible 
to use a single set of reading amplifiers to read from one of four differ- 
ent groups of heads where recording was going on simultaneously in one 
or more of the other groups. To get around this problem it was agreed 
that they would provide two sets of reading amplifiers. Since it seems 
that we will also v*-ant to be reading from other groups of heads for 
functions not involving the four groups raentionec above, they are going 
to provide a third set of amplifiers which will be tied in with the 
group selector matrix for switching hy the computer. To take care of 
equipment which may be reading from the drum entirely independently of 
the computer, a fourth set of reading amplifiers appears to be necessary. 

Of the total of 64. reading amplifiers only 32 will be needed 
initially. Therefore they will plan to include 32 amplifiers with the 
equipment to be furnished on our present order, and will supply the 
additional 32 amplifiers upon receipt of a separate order. ' 

5- Inrfut and Outrut Registers . Since we cannot foresee any need 
forjbhe input and output registers which we called for in the original 
proppsal, it was agreed that these registers would be omitted. This 
partially balances the additional circuitry called for in the reading 

On the auxiliary drum two slight changes werp agreed upon. 
They, are going to separate the storage address register so that the left 
11 digits can be cleared independently of the right' A digits. They also 


Meraorandwfl M-13A7 

are going to insert a 2 microsecono delay into the coincicience pulse 
line. This delay will permit a saving of about 6 microseconcG in the 
read-pulse timing, A 2-microseconcs value vras chosen because they con- 
sider their flip flops require about this length of time to switch. 


Several of the details of their circuits were discussed and 
the following points are significant: 

!• Pulse Length . They would like to use 0.5 microsecond pulses 
in their system instead of 0.1-0.3 microsecond pulses. It was learned 
that their flip flops will trigger satisfactorily on 0.1 microsecond 
pulses so it appeared that there will be no reason for them to retain the 
short pulses within their equipment. TherefojTe it was agreed for them 
to go ahead with the 0.5 microsecono pulse design with the understanding 
that we may have to standardize a few of the control pulses which pass 
between the drum system and the computer to obtain proper operation. 
The number of places where pulses ^11 have to be lengthened or shortened 
should not be very large. 

2. Single Wire Teiminals . In a letter from Butler it had been 
stated that they intend to use "single wire terminals" for connections 
between the drums and the computer. This 'J/^iiainology refers to the way 
in which flip flops are set and means that inputs are provided only to 
the "1" side of the flip flops. Before a number can be read into a 
register the register must first be cleared. This is the same method 
which we use in VA'yI. 

3* Magnetic Heads . There is a duty factor limitation in widting 
with magnetic heads which is imposed by heating v/ithin the cores. For 
the metallic-core heads writing should be limiteo to an average of 16 
microseconds between pulses* With the ferrite heads they felt there 
might also be a similar limitation! but as yet this has to be determined 

Transients produced in the reading amplifier as a result of 
switching between heads or of writing with a head disable the reading 
circuits for various lengths of time. If a head is to be used for reading 
following a writing operation, a minimum of 50 microseconds and preferably 
64 microseconds is required for the transient to die out. If a head is 
to be used for reading foHovdng a switching operation a minimum of 16 
microseconds and preferably about 30 should be provided for the transient 
to disappear. Ip switching from reading to recording, however, no delay 
is required so that a recorc^ing can take plaqe on the next time pulse 
or 8 microseconds after swi"^ching. 


Meinorandum M-13^7 

4.« I/ual Heads . The dual heads that will be iXirziishec vjlth our 
equipment -will be spaced ^ 32 drum registers apart. Adjua-bment of the 
spacing betweeiji these heads vdll allov; a change in separation by ^1 
register. Thege heads are so designed that different spacings could bo 
obtained if desired.. The mdnimum separation is about 25 registers and 
mexamuin. is .79 registers* This separation is controlled by the position 
at which tile head casings ere soldered on to the holder. They have not 
yet fully decided whuther the two heads can be used interchangeably for 
reading and writing. At present they think this will be the case. 

The buff er drum will be provided with 12 dual heads* If one 
wished a drum with all dual heads their present design v;ould pennit A 
such heads jmr axial inch, making a total of 56 heads on the 14" di-um. 
They are making studies of the use of more then one head per track and 
have obtained successful operation of circulating registers in the lab- 
oratory. They are promoting the idea of using the team "revolver" to 
describe a circulating register. 

The dual heads which will be used for the auxiliary marker 
channels on the puffer drum require a continuous erase. At present 
they propose to do this by means of a permanent magnet. 

5« Scares. They are planning to provide spare parts for the 
drum equipment and have included this in their original contract price. 
These spares will be complete plug-in chasses and there will be a mini- 
mum of 10^ of these spares. In cases where the same plug-in unit is 
found in both drums the spares will be figured on the basis of the total 
number of identical units in the two systems* They expect to have extra 
panel space in each of the two systems* After about September of next 
year they expect to stock standard panels and standard chasses so that 
these could be purchased at any time after that by a special order. 


O'Brien and I examinee their circuits in detail with Butler and 
Eulberg and worl^ed out a plan for providing voltage variation in the 
drum systems f 01? marginal checking purposes. We decided that the power 
wiring to the equipment should be grouped as follows for marginal checkings 

1. Flip flop plates. (2 lines each) 
a* Angular position counter 

b. Storage address register, (left 11 digits) 

c. Group address register, (right /i digits of SAR) 

d. Cqntrol flip flops and angular-coincicence-detector-alarm 
flip flop 



Memorandum M-13i^7 

2, Gate tubes, (l line' each) 

a. Angular-position-counter carry gates • 
b« Storage-addresa-register carry gates 

c. Gates in control circuits and ACL alarm 

d. Output gates on reading amplifier 
e« GL series gates 

3« Reading amplifiers (l line each) 

a* Status head amplifiers 

b. Ausdliary marker amplifiers 

o« Information channel amplifiers 
d. Timing track arai)llfier 
e« Bracket track amplifier 

A* Writing amplifiers, (l line each) 

a. Status track Tn?iters 

b. Auxiliary marker writers 
c» Information track v/riters 

The above breakdowi of marginal checking requires a total of 
21 voltage variation lines. It may be true that this breakdown is finer 
than is necessary but if so there will be no problem in combining some 
of these lines at a later date to reduce the number of voltage variation 
circuitso v 


E. S. Rich 


cc: H« Fahnestock 
R. R. Everel^t 
N, H. Taylor 
C. R. Wieser 
J. A. O'Bidep 
R. L. Vk'alquist 

B. E. Morrias Jr. 
S. H. Lodd JV. 

C. W. Watt Jr. 
H. B. Morley